Original Paper Urol Int 1992:48:171-174
Department of Urology, University Hospital of Patras, Greece
Key Words Prostate specific antigen Prostate cancer
Prostate-Specific Antigen Test: Operating Characteristics and Assessment Criteria in the Diagnosis of Prostatic Cancer Abstract The sensitivity and specificity of prostate-specific antigen (PSA) in prostatic carcinoma is of considerable interest. In this study, we have assessed PSA and also correlations between positive and negative predictive values of PSA and the prevalence of prostatic cancer. Firstly, cutoff point must be selected as a positivity criterion for prostatic cancer, according to the purpose of the test. Besides sensitivity, all the other operating characteristics are dependent upon the prevalence of the disease. Specificity and positive predictive value increase while negative predictve value decreases when prevalence increases. As the relationship between the age and the prevalence of prostatic carcinoma is well known, the age of the patient becomes a variable of prominent impor tance when assessing the test.
Introduction In 1979, under the auspices of the National Prostate Cancer Project, a new biological marker for the prostatic tissue was discovered, the prostate-specific antigen (PSA) [1], Nadji et al. [2] and Papsidero et al. [3] have shown that only prostatic epithelial cells produce PSA. There fore, it is possible to apply the PSA as a tumor marker for adenocarcinoma of the prostate. Several researchers have attempted to evaluate the role of PSA in the detection and management of prostatic cancer. In a series of papers, the interest focused mainly on the sensitivity and specificity of PSA [4-6]. Sensitivity and specificity are reliable indi cators of the potential usefulness of a diagnostic test. However, their absolute values vary with the criteria of positivity and negativity of the procedure, the type of the population investigated and the definition of the disease. In this investigation, we reevaluate the role of PSA in the
Received: July 24. 1990 Accepted after revision: March 18. 1991
diagnosis of prostatic cancer by estimating all the operat ing characteristics of the test, showing also the depen dence of these upon the aforementioned statistical param eters.
Materials and Methods Serum samples were studied from 110 patients with a histological diagnosis of prostatic cancer in all stages, who had not been treated and had not undergone any manipulation for at least 5 days prior the test. The mean age was 73.5 (range: 60-93). We also evaluated 78 patients with prostatism who had biopsies or resections which revealed evidence of benign prostatic hypertrophy (BPH). Their mean age was 72.5 (range: 58-87). In addition, 64 males, matched for age, from the general medicine department without urological his tory and symptoms served as controls. We used the reagent Tandem Hybritech by radioimmunoassay, and analyses were performed following the protocol described in the product inserts.
Petros Perimenis, MD Nafpaktou 42 GR-263 31 Patras (Greece)
© 1992 S. Karger AG, Basel (K)42-1138/92/0482-0171 $2.75/0
Downloaded by: King's College London 137.73.144.138 - 3/5/2018 6:59:18 AM
P. Perimenis S.A. Markon G. Barbalias
Table 1. Mean PSA levels and classification according to age Group
PSA levels, ng/ml mean ± SD age groups 50-59 60-69 70-79 80-89 90-99
Normal (n = 64)
l.4±0.3
1.1
1.1
1.3
2.1
BPH (n = 78)
9.8 ±13.8
5.2
5.7
10.9
17.3
Cancer (n = 110)
28 + 20.8
-
19.5
30.1
21.2
Specificity, %
Results The mean PSA levels in our patients are shown in table 1. In order to decide the optimum cutoff point for the PSA test, we constructed a characteristic curve (fig. 1). The ideal cutoff point is one in which no overlap is observed in the range of results among patients with and without prostatic cancer. However, some overlap of find ings is seen among those with BPH and those with pros tatic cancer. We have chosen the PSA level of 9.8 ng/ml as critical for our purposes. This choice results in high specificity, 89%, but without limited sensitivity, 71 %. Such a cutoff point may be appropriate to confirm a suspected diagno sis of prostatic cancer and also can be used to screen for or to exclude disease because of the high sensitivity. It is known that a negative biopsy for prostatic cancer does not exclude it, or, on the other hand, the histologic diagnosis of BPH does not rule out cancer in other sec tions of prostatic tissue which were not examined. This results in an inaccurate determination of the true status of the patients and causes a misclassification of individuals
172
Fig. 1. Characteristic curve of the PSA test. Numbers represent concentrations of PSA (ng/ml; cuttoff points). The thick arrow indi cates our selected cutoff point.
in the group without prostatic cancer. This statistical phe nomenon was further elaborated by Staquet et al. [10], who have shown that only the sensitivity of the test can be correctly calculated. Specificity, positive and negative predictive values are variable because they are influenced by the prevalence of the disease. These investigators cal culated exactly the dependence of the aforementioned test characteristics upon the prevalence of the disease. In fig ures 2 and 3, specificity, positive and negative predictive values of the PSA test (cutoff point: 9.8 ng/ml) are depicted graphically according to the prevalence of pros tatic cancer in the population tested. For our group we found positive and negative predicitve values of 83 and 79%, respectively.
Discussion It has been clearly shown that the proper selection and interpretation of diagnostic tests and procedures can be guided by the principles of diagnostic reasoning and knowledge of test characteristics [11], PSA is a new biological marker, and its measurement in serum has been accepted as one of the tests used to diagnose and monitor prostatic cancer. Our study does not aim to determine the operating characteristics and the diagnostic value of the test, especially for every stage sepa rately, as several researchers have computed and pub lished some of them since its discovery. With the help of these characteristics, we try to indicate an additional way
Perimenis/Markou/Barbalias
PSA: Operating Characteristics and Assessment Criteria
Downloaded by: King's College London 137.73.144.138 - 3/5/2018 6:59:18 AM
Statistics Determined by standard methods [7], sensitivity is the percent age of patients with prostatic cancer whose test values exceed the decision value, and specificity is the percentage of patients without cancer whose test values are less than or equal to the decision value. The positive predictive value is the probability of cancer exis tence when the test is positive, and the negative predictive value is the probability of cancer absence when the test is negative. The cutoff point is the arbitrary value used to separate positive from negative test results and the curve is a graphic plot of sensitivity versus speci ficity at various cutoff points. We use the term prevalence not in its strict epidemiological meaning but, as other authors have, as the per centage of patients with prostatic cancer in the population tested [8, 9],
100-1
o 'o
g. 40to
200 -|----------------- 1------------------- 1-------------------- 1------------------- 1-------------------- 1
10
20 30 Prevalence
40
50 Prevalence
Fig. 2. Specificity (%) of PSA test according to the prevalence of prostatic carcinoma, (cutoff point: 9.8 ng/ml).
Fig. 3. Positive (+) and negative (-) predictive values of the PSA test for the diagnosis of prostatic cancer according to the prevalence of the disease (cutoff point: 9.8 ng/ml).
for the consideration of PSA according to the purpose of the test and the age of the patients. Thus we have first constructed a characteristic curve, which permits to the individual clinician to decide the optimum cutoff point in the light of special circumstances. When the point used to define an abnormal result is situated in the direction of patients with prostatic carcinoma, specificity increases, but no significantly, and sensitivity decreases. When it is moved in the direction of patients without prostatic carci noma, the reverse happens. As referred, we have selected a PSA level of 9.8 ng/ml with 71% sensitivity and 89% specificity. For screening, a much higher sensitivity is required, but the process of confirming prostatic carci noma obviously requires high specificity. So we believe that practically the concept of decision values is more proper than absolute definitions of abnormality and nor mality. The correct evaluation of a new diagnostic test requires the knowledge of the true status of the patient. In our patients, the status was determined through biopsy of the prostate or by means of pathologic evaluation of tissue removed by transurethral or transvesical prostatectomy, but these reference tests however are imperfect; while their sensitivity is 100%, their specificity is unknown. It follows that a few patients with prostatic cancer were probably classified incorrectly as BPH. Thus only sensi tivity is correctly evaluated, while specificity, positive and negative predictive values are proportional to the preva lence of prostatic cancer in the population tested. As it is shown, specificity and positive predictive value increase
while prevalence increases. In case of the negative predic tive value, a decrease is observed while prevalence in creases. In a recent review [9], the incidence rate of prostatic carcinoma varies with age, increasing from 6.3% in the 5th decade to 60% in the 10th decade of life. For the clini cal practice, this relation and the referred results mean that the patient’s age is an important parameter to be con sidered in the evaluation of the PSA test for diagnosis of prostatic cancer. Older men with prostatism and high PSA levels in serum should be given priority on the wait ing list because of the increased possibility of cancer exis tence. Furthermore, the PSA test may be useful as a part of screening for prostatic cancer in combination with dig ital examination. The latter increases the prevalence of the disease in the population tested by further selecting individuals suspicious of having cancer. PSA test and dig ital examination represent important diagnostic parame ters, and the probable inclusion of other tests may lead to a more complete screening of patients or diagnostic evalu ation of cancerous individuals.
173
Downloaded by: King's College London 137.73.144.138 - 3/5/2018 6:59:18 AM
0
References
174
5 Guinan P. Bhatti Ray P: An evaluation of pros tate specific antigen in prostatic cancer. J Urol 1987;137:686-689. 6 Seamonds B, Yang N, Anderson K. et al: Eval uation of prostate specific antigen and pros tatic acid phosphatase as prostatic cancer markers. Urology 1986:28:472-479. 7 Griner PF, Maycwski RJ. Mushlin AI. et al: Selection and interpretation of diagnostic tests and procedures, principles and application. Ann Intern Med 1981;94:553-600. 8 Rullis I. Shaeffer JA, Lilien OM: Incidence of prostatic carcinoma in the elderly. Urology 1975;6:295-297.
Perimenis/Markou/Barbalias
9 Sheldon CA. Williams RD. Fraley EE: Inciden tal carcinoma of the prostate: A review of the literature and critical reappraisal of classifica tion. J Urol 1980:124:626-631. 10 Staquet M, RozcncweigM, Lee YJ. et al: Meth odology for the assessment of new dichoto mous diagnostic tests. J Chronic Dis 1981:34: 599-610. 11 Galea RS, Gambino SR: Beyond Normality: The Predicitve Value and Efficiency of Medi cal Diagnoses. New York. Wiley. 1975.
PSA: Operating Characteristics and Assessment Criteria
Downloaded by: King's College London 137.73.144.138 - 3/5/2018 6:59:18 AM
1 Wang MC, Valenzuela LA. Murphy GP, et al: Purification of human prostate specific anti gen. Invest Urol 1979;17:159-163. 2 Nadji M. Tabei SZ. Castro A, et al: Prostate specific antigen: An immunohistologic marker for prostatic neoplasms. Cancer 1981 ;48: 1229-1232. 3 Papsidero LD. Kuriyuama M, Wang CM. ct al: Prostate antigen: A marker for human prostate epithelial cells. J.N.C.I. 1981:66:37-41. 4 Emtage LA, Lewis PW, Blackledge GRP: The role of prostatic specific antigen in the baseline assessment of patients undergoing hormone therapy for advanced prostate cancer. Br J Urol 1987:60:572-577.
Department of Urology, University Hospital of Patras, Greece
Key Words Prostate specific antigen Prostate cancer
Prostate-Specific Antigen Test: Operating Characteristics and Assessment Criteria in the Diagnosis of Prostatic Cancer Abstract The sensitivity and specificity of prostate-specific antigen (PSA) in prostatic carcinoma is of considerable interest. In this study, we have assessed PSA and also correlations between positive and negative predictive values of PSA and the prevalence of prostatic cancer. Firstly, cutoff point must be selected as a positivity criterion for prostatic cancer, according to the purpose of the test. Besides sensitivity, all the other operating characteristics are dependent upon the prevalence of the disease. Specificity and positive predictive value increase while negative predictve value decreases when prevalence increases. As the relationship between the age and the prevalence of prostatic carcinoma is well known, the age of the patient becomes a variable of prominent impor tance when assessing the test.
Introduction In 1979, under the auspices of the National Prostate Cancer Project, a new biological marker for the prostatic tissue was discovered, the prostate-specific antigen (PSA) [1], Nadji et al. [2] and Papsidero et al. [3] have shown that only prostatic epithelial cells produce PSA. There fore, it is possible to apply the PSA as a tumor marker for adenocarcinoma of the prostate. Several researchers have attempted to evaluate the role of PSA in the detection and management of prostatic cancer. In a series of papers, the interest focused mainly on the sensitivity and specificity of PSA [4-6]. Sensitivity and specificity are reliable indi cators of the potential usefulness of a diagnostic test. However, their absolute values vary with the criteria of positivity and negativity of the procedure, the type of the population investigated and the definition of the disease. In this investigation, we reevaluate the role of PSA in the
Received: July 24. 1990 Accepted after revision: March 18. 1991
diagnosis of prostatic cancer by estimating all the operat ing characteristics of the test, showing also the depen dence of these upon the aforementioned statistical param eters.
Materials and Methods Serum samples were studied from 110 patients with a histological diagnosis of prostatic cancer in all stages, who had not been treated and had not undergone any manipulation for at least 5 days prior the test. The mean age was 73.5 (range: 60-93). We also evaluated 78 patients with prostatism who had biopsies or resections which revealed evidence of benign prostatic hypertrophy (BPH). Their mean age was 72.5 (range: 58-87). In addition, 64 males, matched for age, from the general medicine department without urological his tory and symptoms served as controls. We used the reagent Tandem Hybritech by radioimmunoassay, and analyses were performed following the protocol described in the product inserts.
Petros Perimenis, MD Nafpaktou 42 GR-263 31 Patras (Greece)
© 1992 S. Karger AG, Basel (K)42-1138/92/0482-0171 $2.75/0
Downloaded by: King's College London 137.73.144.138 - 3/5/2018 6:59:18 AM
P. Perimenis S.A. Markon G. Barbalias
Table 1. Mean PSA levels and classification according to age Group
PSA levels, ng/ml mean ± SD age groups 50-59 60-69 70-79 80-89 90-99
Normal (n = 64)
l.4±0.3
1.1
1.1
1.3
2.1
BPH (n = 78)
9.8 ±13.8
5.2
5.7
10.9
17.3
Cancer (n = 110)
28 + 20.8
-
19.5
30.1
21.2
Specificity, %
Results The mean PSA levels in our patients are shown in table 1. In order to decide the optimum cutoff point for the PSA test, we constructed a characteristic curve (fig. 1). The ideal cutoff point is one in which no overlap is observed in the range of results among patients with and without prostatic cancer. However, some overlap of find ings is seen among those with BPH and those with pros tatic cancer. We have chosen the PSA level of 9.8 ng/ml as critical for our purposes. This choice results in high specificity, 89%, but without limited sensitivity, 71 %. Such a cutoff point may be appropriate to confirm a suspected diagno sis of prostatic cancer and also can be used to screen for or to exclude disease because of the high sensitivity. It is known that a negative biopsy for prostatic cancer does not exclude it, or, on the other hand, the histologic diagnosis of BPH does not rule out cancer in other sec tions of prostatic tissue which were not examined. This results in an inaccurate determination of the true status of the patients and causes a misclassification of individuals
172
Fig. 1. Characteristic curve of the PSA test. Numbers represent concentrations of PSA (ng/ml; cuttoff points). The thick arrow indi cates our selected cutoff point.
in the group without prostatic cancer. This statistical phe nomenon was further elaborated by Staquet et al. [10], who have shown that only the sensitivity of the test can be correctly calculated. Specificity, positive and negative predictive values are variable because they are influenced by the prevalence of the disease. These investigators cal culated exactly the dependence of the aforementioned test characteristics upon the prevalence of the disease. In fig ures 2 and 3, specificity, positive and negative predictive values of the PSA test (cutoff point: 9.8 ng/ml) are depicted graphically according to the prevalence of pros tatic cancer in the population tested. For our group we found positive and negative predicitve values of 83 and 79%, respectively.
Discussion It has been clearly shown that the proper selection and interpretation of diagnostic tests and procedures can be guided by the principles of diagnostic reasoning and knowledge of test characteristics [11], PSA is a new biological marker, and its measurement in serum has been accepted as one of the tests used to diagnose and monitor prostatic cancer. Our study does not aim to determine the operating characteristics and the diagnostic value of the test, especially for every stage sepa rately, as several researchers have computed and pub lished some of them since its discovery. With the help of these characteristics, we try to indicate an additional way
Perimenis/Markou/Barbalias
PSA: Operating Characteristics and Assessment Criteria
Downloaded by: King's College London 137.73.144.138 - 3/5/2018 6:59:18 AM
Statistics Determined by standard methods [7], sensitivity is the percent age of patients with prostatic cancer whose test values exceed the decision value, and specificity is the percentage of patients without cancer whose test values are less than or equal to the decision value. The positive predictive value is the probability of cancer exis tence when the test is positive, and the negative predictive value is the probability of cancer absence when the test is negative. The cutoff point is the arbitrary value used to separate positive from negative test results and the curve is a graphic plot of sensitivity versus speci ficity at various cutoff points. We use the term prevalence not in its strict epidemiological meaning but, as other authors have, as the per centage of patients with prostatic cancer in the population tested [8, 9],
100-1
o 'o
g. 40to
200 -|----------------- 1------------------- 1-------------------- 1------------------- 1-------------------- 1
10
20 30 Prevalence
40
50 Prevalence
Fig. 2. Specificity (%) of PSA test according to the prevalence of prostatic carcinoma, (cutoff point: 9.8 ng/ml).
Fig. 3. Positive (+) and negative (-) predictive values of the PSA test for the diagnosis of prostatic cancer according to the prevalence of the disease (cutoff point: 9.8 ng/ml).
for the consideration of PSA according to the purpose of the test and the age of the patients. Thus we have first constructed a characteristic curve, which permits to the individual clinician to decide the optimum cutoff point in the light of special circumstances. When the point used to define an abnormal result is situated in the direction of patients with prostatic carcinoma, specificity increases, but no significantly, and sensitivity decreases. When it is moved in the direction of patients without prostatic carci noma, the reverse happens. As referred, we have selected a PSA level of 9.8 ng/ml with 71% sensitivity and 89% specificity. For screening, a much higher sensitivity is required, but the process of confirming prostatic carci noma obviously requires high specificity. So we believe that practically the concept of decision values is more proper than absolute definitions of abnormality and nor mality. The correct evaluation of a new diagnostic test requires the knowledge of the true status of the patient. In our patients, the status was determined through biopsy of the prostate or by means of pathologic evaluation of tissue removed by transurethral or transvesical prostatectomy, but these reference tests however are imperfect; while their sensitivity is 100%, their specificity is unknown. It follows that a few patients with prostatic cancer were probably classified incorrectly as BPH. Thus only sensi tivity is correctly evaluated, while specificity, positive and negative predictive values are proportional to the preva lence of prostatic cancer in the population tested. As it is shown, specificity and positive predictive value increase
while prevalence increases. In case of the negative predic tive value, a decrease is observed while prevalence in creases. In a recent review [9], the incidence rate of prostatic carcinoma varies with age, increasing from 6.3% in the 5th decade to 60% in the 10th decade of life. For the clini cal practice, this relation and the referred results mean that the patient’s age is an important parameter to be con sidered in the evaluation of the PSA test for diagnosis of prostatic cancer. Older men with prostatism and high PSA levels in serum should be given priority on the wait ing list because of the increased possibility of cancer exis tence. Furthermore, the PSA test may be useful as a part of screening for prostatic cancer in combination with dig ital examination. The latter increases the prevalence of the disease in the population tested by further selecting individuals suspicious of having cancer. PSA test and dig ital examination represent important diagnostic parame ters, and the probable inclusion of other tests may lead to a more complete screening of patients or diagnostic evalu ation of cancerous individuals.
173
Downloaded by: King's College London 137.73.144.138 - 3/5/2018 6:59:18 AM
0
References
174
5 Guinan P. Bhatti Ray P: An evaluation of pros tate specific antigen in prostatic cancer. J Urol 1987;137:686-689. 6 Seamonds B, Yang N, Anderson K. et al: Eval uation of prostate specific antigen and pros tatic acid phosphatase as prostatic cancer markers. Urology 1986:28:472-479. 7 Griner PF, Maycwski RJ. Mushlin AI. et al: Selection and interpretation of diagnostic tests and procedures, principles and application. Ann Intern Med 1981;94:553-600. 8 Rullis I. Shaeffer JA, Lilien OM: Incidence of prostatic carcinoma in the elderly. Urology 1975;6:295-297.
Perimenis/Markou/Barbalias
9 Sheldon CA. Williams RD. Fraley EE: Inciden tal carcinoma of the prostate: A review of the literature and critical reappraisal of classifica tion. J Urol 1980:124:626-631. 10 Staquet M, RozcncweigM, Lee YJ. et al: Meth odology for the assessment of new dichoto mous diagnostic tests. J Chronic Dis 1981:34: 599-610. 11 Galea RS, Gambino SR: Beyond Normality: The Predicitve Value and Efficiency of Medi cal Diagnoses. New York. Wiley. 1975.
PSA: Operating Characteristics and Assessment Criteria
Downloaded by: King's College London 137.73.144.138 - 3/5/2018 6:59:18 AM
1 Wang MC, Valenzuela LA. Murphy GP, et al: Purification of human prostate specific anti gen. Invest Urol 1979;17:159-163. 2 Nadji M. Tabei SZ. Castro A, et al: Prostate specific antigen: An immunohistologic marker for prostatic neoplasms. Cancer 1981 ;48: 1229-1232. 3 Papsidero LD. Kuriyuama M, Wang CM. ct al: Prostate antigen: A marker for human prostate epithelial cells. J.N.C.I. 1981:66:37-41. 4 Emtage LA, Lewis PW, Blackledge GRP: The role of prostatic specific antigen in the baseline assessment of patients undergoing hormone therapy for advanced prostate cancer. Br J Urol 1987:60:572-577.
